Explanation of Experimental Conditions

The driving environment is focused on highway with different light conditions.

The image sequences captured by the camera with unknown two pan- or tilt-angles are tested with the same lane detection algorithm in Fig. 5-3. At the same time, in order to observe if the lane-based warning system can maintain robust performance and tolerate the light variation, we select the video segments with three different periods, daytime, evening, and night of one day for experiment in the next section.

Fig. 5-3 : The testing image with different mounting angles.

5.2.2 Results of Lane Det ction

In Fig. 5-4, the testing environment considers the two properties with respect to the

e

view-angles and light conditions simultaneously. The detection results of daytime (a), evening (b), and night (c) are processed by the same programming setting.

Fig. 5-4 : The results of lane detection.

In Fig. 5-4 (c), the lane boundary can be clearly extracted in the nighttime driving environment even if the side-view of vehicle usually has more chances to subject to the perturbation from the exterior light-sources.

5.2.3 Results of Lane Departure Warning

If the lane boundary is locked precisely by the lane detection mechanism, the lane departing maneuver can be tracked and recorded its position whether the lateral speed is faster or not. Figure 5-5, 5-6, and 5-7 shows the tracking results of lane departure with different variations of light and moving direction of the vehicle.

# Frame 3293 # Frame 3299

# Frame 3305 # Frame 3310

# Frame 3316 # Frame 3334

# Frame 3385 # Frame 3388

Fig. 5-5 : The results of lane departure caused by cutting into the inside lane.

# Frame 693 # Frame 718

# Frame 733 # Frame 746

# Frame 787 # Frame 818 Fig. 5-6 : The results of lane departure in the night time.

# Frame 3689 # Frame 3701

# Frame 3740 # Frame 3776

Fig. 5-7 : The results of lane departure caused by moving into the outside lane.

5.2.4 Results of Stable-Driving Region and Drowsiness Estimation System

As described in Section 4.3.3, the straight-road driving distance between the lane marker and wheels can be modeled by the clustered distribution with higher weight and smaller standard deviation. For further adaptation, we develop an update mechanism to make the stable region adaptive to the changeful driving habits of people. Figure 5-8 shows the updating process of stable-region described as a statistical chart which contains the information of lateral offsets at the same time.

From Fig. 5-8 (a) to (d), the mean value of the stable-region will increase obviously due to the accumulated lateral offsets which are almost situated over the region and can be regarded as the new driving habit of the driver adequately.

(a) (b)

(c) (d)

Fig. 5-8 : Results of update for the stable-driving region.

(a) (b)

(c) (d)

Fig. 5-9 : Results of the variation of drivers’ drowsy degree by the reaction time.

The relationship between the gauge chart of drowsy degree and the reaction time of drivers is demonstrated in Fig. 5-9. From Fig. 5-9 (a) to (b), the reaction time will start to be counted since the lateral offset is outside the stable-region at that moment.

Therefore, the drowsy degree can be raised with a specific ratio of the measured reaction time to the threshold which has been evaluated by the EGG-based analysis from BRC. On the other hand, from Fig. 5-9 (c) to (d), the drowsy degree keeps increasing because the time interval between the current and previous reaction time which are both greater than the threshold is not for 10 sec.

5.3 Performance

Table 6 : The processing information of lane detection algorithms.

Output

Fig. 5-10 : The processing ratio of 4 examples implemented by lane detection and lane departure warning algorithms.

The performance of the four testing videos with two different view-angles is listed in Table 6. We split the lane detection approach into three parts. (A): The de-noise preprocessing. (B): The lane boundary extraction. (C): The edge linking task and LDW. In general, the size of ROI has the most influence on the systematic execution time. However, the frequently departing maneuver occurred in the image

segments will result in the additional computing load. As explained in Section 3.4, the bottom sub-region of ROI must be searched for line-pixels with Hough Transform for each frame. In other words, the searching range in it can not be regarded as the limited size near the position determined by the previous lane marker if the lateral offsets change seriously. Figure 5-10 shows that the ratio of edge linking mechanism depends on the image contents whether the ROI is large or not. On the other hand, due to the 5x5 size of Gaussian mask, the de-noise procedure still occupies most of execution time even if the frequent departure exists or not.

5.4 Discussion and Analysis

In order to increase detection rate of LDW and drowsiness estimation system, the lane detecting error must be low as much as possible even if this algorithm is always subjected to the disturbance resulted from external factors. In Fig. 5-11, the lane markers can be still extracted by our developing method although they are unclear.

However, the lane detection method we proposed in this thesis can not resolved some cases such as driving in a tunnel so that the contrast between lane markers and road surface is not enough, as shown in Fig. 5-12 (a).

Fig. 5-11 : Results of lane detection for the unclear lane markers.

(a) (b)

(c)

Fig. 5-12 : Some examples of detecting error in our lane detection system.

In addition, as explained in Section 3.2, the range of ROI can be detected by the boundary information of the car window and that of the horizon in the image. But this property may be not suited to the environment which does not only contain the above clues for ROI extraction but be affected by the light conditions, such as the example shown in Fig. 5-12 (b). On the other hand, the external light in the nighttime has chance to produce a “light ring” effect on the camera lens which may cause the deviated parting position of the detected lane markers in the image instantaneously, as shown in Fig. 5-12 (c).

6 Chapter 6 Conclusions

We propose a scientific system for driver’s drowsiness estimation by integrating the statistics evaluated by the EEG-based analysis from BRC into our lane departure warning system in the realistic driving environment. In this thesis of lane detection, we develop a method for automatic ROI extraction only by analyzing the image contents captured by the fish-eye camera mounted under the rear-view mirror without knowing the related camera parameters in advance. To overcome the light variations, the de-noising architecture which is considered the spatial and temporal domain at the same time can restrain the noise effectively. Focused on the geometric property of the blind-spot view, the adaptive type of edge operator and threshold selection can exactly detect the lane boundary. Finally, an improved edge linking model we proposed not only increases the searching speed for lane trajectory but resolves the effect of fish-eye lens distortion.

About the lane departure warning and drowsiness estimation proposed in this thesis, we construct a warning mechanism with the lateral offsets and TLC computed by the lateral velocity and the border of ROI. Due to the different driving habits of people, we construct the stable-driving region for modeling by the information of previous lateral positions of lane markers with updating mechanism. Then, we use the deviation as the index for drowsiness estimation which has been analyzed and evaluated by EEG-analysis. By considering the human’s behavioral style that the reactive behavior must be increasingly slower for a long period when the subjects enter the drowsy state gradually, we design a gauge of drowsy degree to estimate the driver’s psychological state according to the reaction time of drivers.

The lane-based stable system has been tested that the average frame rate is up to 15fps on PC platform. In the future, it will be integrated into the blind-spot side collision warning architecture to increase the better detection rate and provide more adaptive performance. Besides, by the constructed mechanism for drowsiness estimation in the dynamic driving environments, we can collect more data to further analyze the other inattentive behavior of drivers through this system so that the safety driving system can consider all the possible risks caused by the internal or external factors of drivers as much as possible.

References

[1] G. Jacobs, A. Aeron-Thomas, and A. Astrop, “Estimating global road fatalities,”

Australian National University, Transport Research Laboratory, Technical Report TRL 445, 1999.

[2] “Traffic safety facts 2004.” Technical report, National Center for Statistics and Analysis, U.S. Department of Transportation, 2004.

[3] http://www.motc.gov.tw/hypage.cgi?HYPAGE=index.htm

[4] A. Broggi, M. Betozzi, A. Fascioli, C. Guarino, L. Bianco, and A. Piazzi, “The argo autonomous vehicle’s vision and control systems,” Int’l Journal Intelligent Control and Systems, vol. 3, no. 4, pp. 409-441, 1999.

[5] M. Bertozzi and A. Broggi, “GOLD: a parallel real-time stereo vision system for generic obstacle and lane detection,” IEEE Trans. Image Processing, vol. 7, no. 1, Jan. 1998.

[6] G. Y. Jiang, T. Y. Choi, S. K. Hong ,J. W. Bae, and B. S. Song, “Lane and obstacle detection based on fast inverse perspective mapping algorithm,” in IEEE Conf. Systems, Man, and Cybernetics, vol. 4, pp. 2969-2974, Oct. 2000.

[7] Y. Wang, D. Shen, and E. K. Teoh, “Lane detection using Catmull-Rom Spline,”

in Proc. IEEE Intelligent Vehicles Symp., Stuttgart, Germany, pp. 51-57, 1998.

[8] Y. Wang, E. K. Teoh, and D. Shen, “Lane detection and tracking using B-Snake,”

Image and Vision Computing, vol. 22, pp. 269-280, 2004.

[9] C. R. Jung and C. R. Kelber, “A robust linear-parabolic model for lane following,” in Proc. SIBGRAOI, Curitiba, PR, pp. 72-79, Oct. 2004

[10] C. C. Wang, S. S. Huang, and L. C. Fu, “Driver assistance system for lane detection and vehicle recognition with night vision”, in Proc. IEEE Intelligent Robots and Systems, pp. 3530-3535, Aug. 2005.

[11] R. Risack, N. Mohler, and W. Enkelmann, “A video-based lane keeping assistant”, in Proc. IEEE Intelligent Vehicles Symp., Dearborn, MI, pp. 506-511, Oct. 2000.

[12] S. Y. Kim and S. Y. Oh, “A driver adaptive lane departure warning system based on image processing and a fuzzy evolutionary technique,” in Proc. IEEE Intelligent Vehicles Symp., Columbus, OH, pp.361-365, Jun. 2003.

[13] K. Weiss, N. Kaempchen, and A. Kirchner, “Multiple-model tracking for the detection of lane change maneuvers,” in Proc. IEEE Intelligent Vehicles Symp., Parma, Italy, pp. 937-942, 2004.

[14] W. Enkelmann, “Video-based driver assistance-from basic functions to applications,” in Proc. Int’l Journal Computer Vision, vol. 45, no. 3, pp.

201-221 , Dec. 2001.

[15] P. H. Batavia, “Driver-adaptive lane departure warning systems,” doctoral dissertation, Robotics Institute, Carnegie Mellon University, Sep. 1999.

[16] R. Kovordanyi, “Dynamically deployed support as a potential solution to negative behavior adaptation,” in Proc. IEEE Intelligent Vehicles Symposium, pp.

613-618, 2005

[17] 林亨杰, ”高速公路速限提昇前後之交通事故探討,＂道路交通安全與執法

國際研討會, Sep. 2004.

[18] M. J. Flannagan, “Current status and future prospects for non-planar rearview mirrors” in SAE Technical Paper Series No. 2000-01-0324

[19] 席世民, “學校交通安全教育工作推動現況與基本安全觀念介紹,＂交通

部道路交通安全督導委員會, Aug. 2004

[20] C. Kreucher and S. Lakshmanan, “LANA: A lane extraction algorithm that uses frequency domain features,” IEEE Trans. Robotics and Automation, vol. 15, no.

2, Apr. 1999.

[21] Y.-U. Yim and S.-Y. OH, “Three-feature based automatic lane detection algorithm (TFALDA) for autonomous driving,” IEEE Trans. Intelligent Transportation Systems, Vol. 4, NO. 4, Dec. 2003.

[22] Q. Li, N. Zheng, and H. Cheng, “Springrobot: a prototype autonomous vehicle and its algorithms for lane detection,” IEEE Trans. On Intelligent Transportation Systems, Vol. 5, NO. 4, Dec. 2004.

[23] M. Chen T. Jochem, and D. “AURORA: a vision-based roadway departure warning system,”

, Pomerleau,

in Proc. IEEE Int’l Conf. Intelligent Robots and Systems, vol. 1, pp. 243-248, Aug. 1995

[24] C.-H. Yeh and Y.-H. Chen, “Development of vision-based lane and vehicle detecting systems via the implementation with a dual-core DSP,” in Proc. IEEE Intelligent Transportation Systems Conf., Sep. 17-20, 2006.

[25] R. Chapuis, R. Aufrere, and F. Chausse, “Accurate road following and reconstruction by computer vision,” IEEE Trans. Intelligent Transportation Systems, vol. 3, no. 4, Dec. 2002.

[26] J. W. Lee, “A machine vision system for lane-departure detection,” in Proc.

Computer Vision and Image Understanding, vol. 86, pp. 52-78, Apr. 2002.

[27] C.-T. Lin, R.-C. Wu, S.-F. Liang, W.-H. Chao, Y.-J. Chen, and T.-P. Jung,

“EGG-based drowsiness estimation for safety driving using independent component analysis,” IEEE Trans. Circuits and Systems, vol.52, Dec. 2005.

[28] A. Lopez, C. Canero, J. Serrat, J. Saludes, F. Lumbreras, and T. Graf, “Detection of lane markings based on ridgeness and RANSAC,” in Proc. IEEE Conf. on Intelligent Transportation Systems, pp.254-259, Sep. 13-16, 2005.

[29] C. R. Jung and C. R. Kelber, “A lane departure warning system using lateral offset with uncalibrated camera,” in Proc. IEEE Conf. on Intelligent Transportation Systems, pp. 102-107, Sep. 13-16, 2005.

[30] D. J. Kang, J. W. Choi, and I. S. kweon, “Finding and tracking road lanes using line-snakes”, in Proc. IEEE Conf. on Intelligent Vehicles Symposium, pp.

189-194, Sep. 1996

[31] M. Ruder, W. Enkelmann, R. Garnitz, “Highway lane change assistant,” in Proc.

IEEE Conf. on Intelligent Vehicles Symposium, Vol.1, pp.240-244, Jun. 2002.

[32] C. Stauffer, W.E.L Grimson, “Adaptive background mixture models for real-time tracking,” in Proc. IEEE Computer Society Conference on Computer Vision and Pattern Recognition,” Vol.2, 1999.

[33] R. K. Lenz and R. Y. Tsai, “Techniques for calibration of the scale factor and image center for high accuracy 3-D machine vision metrology,” IEEE Trans. On Pattern Analysis and Machine Intelligence, Vol.10, NO. 5, Sept. 1988.

[34] R.I. Hartley and S. B. Kang, “Parameter-tree radial distortion correction with centre of distortion estimation,” in Proc. IEEE International Conf. On Computer Vision, Vol.2, pp. 1834-1841, Oct. 2005.

[35] H. Godthel, P. Milgram, and G. J. Blaauw, “ The development of a time-related measure to describe driver strategy,” Human Factors, 26(3): 257-268, 1984.

[36] P. Philip, J. Taillard, “Effect of fatigue on performance measured by a driving simulator in automobile drivers,” Journal Psychosomatic Research 55, pp.

197-200, 2003.

[37] J. Hendrix, “Fatal crash rates for tractor-trailers by time of day,” in Proc. Int’l Truck and Bus Safety Research and Policy Symp., pp. 237-250, 2002.